If you attended OFC 2018 in San Diego last month, you may have heard the buzz around advances in coherent technology to help meet growing bandwidth demand. These advances enable higher capacity and reach, and come in the form of newly demonstrated coherent modulation shaping techniques as well as advanced silicon photonics integration techniques. Whether you were very busy (as I was) in back-to-back meetings or did not have the opportunity to attend OFC, this blog post serves to put some perspective on why all the buzz at OFC and also serves as a refresher on the history of coherent technology, review some of the basics, and look at recent trends on how coherent technology is being applied to various markets.
A Brief History
Coherent optical technology was first introduced in long haul applications to overcome fiber impairments that required complex compensation techniques when using direct detection receivers. Leveraging advanced CMOS processing nodes and reduction in design complexity, coherent solutions have moved from long haul to metro and even shorter reach optical interfaces.
With the introduction of Acacia’s CFP-DCO module in 2014, coherent became an even more compelling solution for metro and data center interconnect (DCI) applications due to its pluggable pay-as-you-grow benefits, and integrated DSP design. Today, coherent is moving from metro core to access aggregation networks (Figure 1). Looking forward, the industry is working to standardize coherent solutions for even shorter reach interfaces.
Figure 1. Coherent solutions transitioning to shorter reaches.
This model of new technology adoption in long haul interfaces, followed by a migration to shorter reach applications, has been demonstrated in the industry before: the copper-to-fiber optics transition followed a very similar path starting in the 1980’s. Technologies, such as dense wavelength division multiplexing (DWDM) and forward error correction (FEC), also followed this same pattern. It is anticipated that this trend of shorter reach applications benefiting from initial long-haul technology investment as it applies to coherent will be no different.
Industry organizations such as the Optical Internetworking Forum (OIF), IEEE, and Cable Labs have initiated coherent standardization activities, recognizing the trend towards using coherent for shorter distances. The OIF is defining a coherent standard for DWDM interfaces in DCI applications with reaches up to 120km; CableLabs is defining coherent standards for cable access networks; and IEEE is considering coherent for unamplified applications beyond 10km. All of this standardization activity reinforces the view of coherent moving to shorter reach, high volume applications. As the applications move from 100G to 400G and beyond, it is likely that coherent will be used in even shorter reach interfaces.
Demand for Bandwidth is Driving New Coherent Markets
Today, coherent technology is already being deployed into markets with a wide array of applications ranging from 10’s of kilometers to 1000’s of kilometers. While network operators in each of these markets need to manage network expansion at the lowest cost, differences in network architectures and demands drive a unique set of priorities for each. Some of these key market applications we’ll explore in this blog include: Long Haul, Metro, DCI, Remote PHY Cable Access, 5G, and applications with unamplified interfaces. We’ll discuss these applications and the factors that drive solutions for their optical interconnecting requirements.
100G Long Haul: Many Fiber Spans with Optical Amplifiers
The long haul market, where coherent was first widely adopted, is still a large segment of the coherent ports shipped each year.
Figure 2. Typical Long Haul network.
This market is highly sensitive to performance because longer reach interfaces eliminate the need for additional costly regeneration. A key optical interconnection performance parameter for coherent DWDM long haul networks is optical signal-to-noise ratio (OSNR).
OSNR & Optical Amplifier Refresher
OSNR describes the relative energy of the signal carrying the information to the energy of the noise from other sources. Optical receivers are specified based on their ability to detect the desired signal in the presence of noise.
So, why is OSNR so critical to long haul DWDM networks? Optical signals are amplified as they are transmitted across the network. At each amplifier, the signal is degraded slightly–the level of the noise is amplified relative to the signal. When the signal level is so close to the noise that it is nearly un-detectable, it needs to be regenerated—the optical signal is converted to the electrical domain where the data can be accurately decoded. The same data is then recoded, retimed, and converted back to the optical domain with a high signal to noise ratio capable of passing through additional amplifiers.
Higher performance coherent optical interfaces allow longer reaches without needing regeneration, which ultimately lowers the cost of deploying and managing these networks.
Currently, 100G QPSK modulation is still widely used in long haul networks because it offers exceptional performance due to achievable OSNR margins and maturity of technology based on 25G electronics. Fiber capacity can be further enhanced by reducing channel spacing from 50GHz to 37.5GHz. Alternatively, capacity can also be increased over the channel by increasing the baud rate. For example, doubling to 200G capacity can be achieved by maintaining QPSK modulation while doubling the baud rate. However, this would require an increase to the channel spacing. To ensure the required channel spacing remains unaltered (e.g., 50GHz), the modulation order also needs to be increased (e.g., from QPSK to 8QAM) as the baud rate is increased.
Simply put, there is an interdependency among baud rate, modulation order, required OSNR, and channel spacing. Transmission capacity can be increased by: (1) increasing the baud rate without changing modulation order, requiring wider channel spacing; (2) holding baud rate constant while moving to a higher modulation order, resulting in narrower channel spacing at the cost of OSNR margin—higher modulation order requires higher OSNR margin; or (3) increasing baud rate and moving to a higher modulation order, requiring no changes to channel spacing but assumes there is sufficient OSNR margin when moving to the higher modulation order. Thus to upgrade to higher capacity with minimal changes to the line system optics, (3) is an attractive option assuming there is sufficient OSNR margin.
Metro: High Density Solutions for ROADM Networks
Coherent interconnections have also been widely adopted in metro core networks (Figure 3). The dynamics in these networks are somewhat different from long haul. Metro core networks usually consist of many reconfigurable optical add-drop multiplexer (ROADM) nodes where wavelengths are dropped, added, or routed to other destinations. This is done using wavelength selective switches (WSSs)–each WSS is effectively an optical filter that can narrow the total bandwidth of the channel. Passing through multiple ROADM nodes can result in significant narrowing of the available spectrum of the complete optical path. These WSSs can introduce additional impairments, such as polarization dependent loss. Though the reaches in these applications are less than long haul, they still require a high level of performance due to the many link impairments.
Metro networks are not entirely about optical performance, though. Central offices can often be crowded and power limited. High density solutions that are power efficient can be particularly valuable in these applications.
Figure 3. Typical Metro network.
In addition, metro networks handle a wide range of traffic. These networks quickly become complex to manage, with different equipment required depending on the requirements. Common solutions that can be leveraged across multiple platforms, supporting a wide variety of traffic types, and scale capacity in a cost effective manner allow network operators to manage their operational costs more effectively.
Data Center Interconnect: Cost and Power Optimized Coherent
In the last 10 years, the optical networking industry has been transformed by the requirements of large cloud network operators. High capacity interconnections are necessary between these hyperscale data centers to enable cloud functionality (Figure 4).
Figure 4. Typical DCI/Cloud network.
These DCI connections differ from traditional carrier networks in that they are usually point-to-point with no ROADMs in between, and can be within the same city or across oceans. Cost and power are generally the most critical parameters for these applications. Some use cases can be fiber constrained, making spectral efficiency a higher priority. Customers in these markets tend to be early adopters of new technology and have short product life cycles.
Large cloud network operators, along with some of the more traditional carriers, are driving changes in how network functions are partitioned between vendors, allowing them greater freedom to transition between vendors and product generations with minimal changes in the software used to control the network.
Remote PHY/Fiber Deeper: Coherent Technology for Cable Access
Access networks are an emerging opportunity for coherent interconnections. The cable industry is taking the lead in this segment by driving standardization of coherent for access aggregation.
Figure 5. Remote PHY network.
As the hybrid fiber-coax (HFC) networks evolve toward remote PHY architectures, fiber is being deployed deeper in the network (Figure 5), resulting in increased available bandwidth to residential end-user customers, while eliminating bottlenecks in the HFC network. 10G optical interfaces are pushed closer to the end users resulting in aggregation points in the network where 10-20 remote PHY devices come together.
Coherent can be an effective way to transport these aggregated signals back to the hub. In some cases, it may only be necessary to transport a single coherent wavelength back to the hub, but since coherent is inherently a DWDM technology, this approach provides the capability to expand capacity by up to two orders of magnitude in the future.
CableLabs is the standards organization of the cable industry and has recognized the need for this solution in the market. In 2017, they kicked off a project to define coherent standards for cable access aggregation applications. Acacia is participating in this project, along with many other leading optical networking vendors, as well as several large MSOs.
5G Drives Backhaul Growth
Another emerging access application is 5G backhaul (Figure 6). It is clear that backhaul demands in wireless networks are going to need to increase significantly. As more capacity is delivered to end users, the connections back to the core network must scale, as well.
Coherent offers several benefits in these access aggregation applications compared to traditional direct detect solutions. At higher data rates, it becomes very challenging to deploy direct detect solutions over 10’s of km’s without using dispersion compensation. Alternatively, solutions may consider many parallel optical interfaces, but that drives up the cost of the solution.
Figure 6. 5G Backhaul network.
Today’s coherent implementations are generally based on tunable laser technology. While fixed laser implementations are a consideration in these applications, tunable solutions can offer operational benefits by significantly reducing the number of spares that need to be stocked. Tunable solutions also tend to support shorter lead times, accelerating the ability to turn up new services. Lastly, coherent solutions are future proof with the ability to scale capacity by increasing data rate or adding additional wavelengths.
Adoption of coherent technology in access networks could offer an additional benefit that may not be obvious at first. Since the same solutions can address a wide range of network interfaces (e.g., access aggregation, metro, and regional), it may be possible to collapse the supply chain for multiple applications into a single solution. This could offer significant operational efficiencies for network operators.
Unamplified Point-to-Point Interfaces
Unamplified point-to-point interfaces are essentially client optical interfaces for connections between buildings (Figure 7). As data rates have increased, it has been more and more challenging for direct detect solutions to address these kinds of applications. At 100G and 200G, proprietary coherent solutions are already used for links in the 40-80km range.
Figure 7. Point-to-point.
Looking forward to 400G, this application is within the scope of the OIF 400ZR project. In addition, the IEEE study group that is considering solutions beyond 10km for data rates of 50G, 100G, 200G, and 400G is evaluating coherent alternatives for these applications.
Since these interconnections are not amplified, they are not characterized by their tolerance to low OSNR. In these applications, transmitter power and receiver sensitivity are the key parameters that define the usable link budget.
Coherent detection offers the same increase in performance for power limited sensitivity as it does for noise limited applications. Volumes for these applications can be larger than transport applications and coherent implementations will need to be cost effective and extremely power efficient.
Coherent is moving to shorter reach as data rates increase
As we’ve outlined in this blog, there are a number of applications for which coherent is well suited. DCI, metro, and long haul are existing markets that have benefited from coherent for several years now. Emerging applications such as Remote PHY for cable access, 5G backhaul, and unamplified “ZR” interfaces are evolving as standards efforts and deployment strategies are still in the early stages. What is clear is that operators struggling to meet the growing demands of bandwidth are motivated to optimize their optical networks for capacity and reach in order to minimize cost. And space and power restrictions continue to be a challenge as additional hardware is deployed in constrained environments. Standardization will help advance the coherent evolution underway.
Stay tuned for upcoming blog posts in which we will focus on how advanced 3D shaping of coherent modulation can optimize various types of coherent networks.